Year 2007
Mikael Louhi

An Electrode Management Model for the Submerged Arc Furnace

The use of artificial intelligence in modelling, controlling and optimizing industrial processes has grown substantially. Numerous successful applications have been reported, especially in the field of metallurgy.

The smelting process of ferrochrome that is performed in a submerged-arc furnace is a multiphase process, and reactions occurring therein are difficult to map. Furnace and electrode functions are constant objects of research. Tools used in the research work include monitoring methods that are based on artificial intelligence.

In the literature study of this thesis a survey of monitoring methods is conducted. In the survey the intelligent monitoring methods: expert systems, fuzzy logics and particularly artificial neural networks, receive extra attention. Theory, architecture and applications of the methods in the metals industry are studied.

In the experimental part of this thesis the ferrochrome smelting process is presented and the functions of and phenomena related to the Söderberg electrodes are scrutinized. The aim of this thesis was to utilize software called CSense to develop a predictive model for the electrode slipping using artificial neural networks. The modelling was conducted with electrode specific process data gathered from a South African ferrochrome plant.

Model development and simulations were based on previous research work and expert knowledge in the field. Of the numerous phenomena related to the electrode functions the following were chosen as modelling variables: electrode current, charge resistance, hydraulic pressure of the slipping device and shift based slipping. Offline simulations were conducted and the obtained results were good. Model prediction accuracies were approximately 60-80% of performed shift based and total slipping. The simulation results lay a good foundation for future studies.

This info last modified 27 Apr 2024 by Jerri Kämpe-Hellenius